A preliminary implementation of metabolic-based pH control to reduce CO2 usage in outdoor flat-panel photobioreactor cultivation of Nannochloropsis oceanica microalgae

Algal Research 18 (2016) 288–295

Contents lists available at ScienceDirect

Algal Research journal homepage: www.elsevier.com/locate/algal

A preliminary implementation of metabolic-based pH control to reduce CO2 usage in outdoor flat-panel photobioreactor cultivation of Nannochloropsis oceanica microalgae Jun Wang a, Theresa Rosov b, Pierre Wensel b, John McGowen b, Wayne R. Curtis a,⁎ a b

Department of Chemical Engineering, The Pennsylvania State University, University Park, PA 16802, United States Arizona Center for Algae Technology and Innovation, Arizona State University, Mesa, AZ 85212, United States

a r t i c l e

i n f o

Article history: Received 15 February 2016 Received in revised form 6 June 2016 Accepted 1 July 2016 Available online xxxx Keywords: pH control Alkalinity Algae cultivation Nannochloropsis oceanic Photobioreactor Nitrogen assimilation Stoichiometry Carbon yield

a b s t r a c t A crucial challenge associated with high-density, commercial-scale, outdoor microalgal cultivation is maintaining pH stability without excessive use of CO2 buffering. This includes a media-dependent, intracellular metabolic proton imbalance leading to the alkalization or acidification of growth media that results when algae consume, respectively, nitrate or ammonium ions. Feeding these two nitrogen sources can theoretically achieve balanced proton metabolism as well as pH control that is economically more favorable as compared to CO2 buffering or acid/base addition. To accomplish this, a fed-batch nutrient feeding strategy must be adopted as a component of a model-based pH control system, particularly in the case of ammonium preference. This work represents a preliminary study of the challenges of implementing a nitrogen metabolism based open-loop pH control strategy in the challenging environment of an outdoor photobioreactor at the DOE-ATP3 testbed facility in Mesa, Arizona during the high solar insolation period of summer 2015. The approach was limited to twice daily fed-batch addition while accounting for ammonium-N preference in a background of nitrate-based algae growth media. Despite these limitations, growth achieved for a photobioreactor operated based on predicted metabolic nitrogen demand (PND) was comparable to ‘CO2-on-demand’ (CoD) for pH control. PND reduced CO2 usage to b10% of CoD control, where the reduced buffering resulted in much greater pH fluctuations due to daily variations in light availability, and a much lower and more consistent media alkalinity (5.73–5.79 mEq/L) as compared to the CoD control (6.51–8.49 mEq/L). While this effort illustrates the utility of feed-forward model-based control, it further illustrates the need for far more sophisticated ‘real time’ monitoring and modeling to accommodate the dynamic outdoor conditions. A need for improved analytics for accurate closure of nitrogen mass balance is also indicated. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Concerns of continued use of transportation fuels and other fossilfuel derivatives in relation to national security, global economics, environment, and rural economic development have led to increased interest in renewable fuels and products from biomass feedstocks like oleaginous microalgae. Nevertheless, commercialization of the algal industry is constrained by high operating costs associated with growth control in large, outdoor raceway open ponds or photobioreactors (PBRs) [13]. Culture pH has the potential to influence a wide range of parameters within an alga production process; extracellular pH influences trans-thylakoidal membrane ionic gradients and electromotive

⁎ Corresponding author. E-mail addresses: [email protected] (J. Wang), [email protected] (T. Rosov), [email protected] (P. Wensel), [email protected] (J. McGowen), [email protected] (W.R. Curtis).

http://dx.doi.org/10.1016/j.algal.2016.07.001 2211-9264/© 2016 Elsevier B.V. All rights reserved.

forces in photosynthesis, the assimilation of nutrients including inorganic carbon, and even the surface charge/zeta-potential that mediate subsequent flocculation-mediated harvesting [9]. Because carbon is typically provided by CO2 (or dissolved wastewater carbon), the major inorganic nutrient (and associated process cost) is nitrogen. Nitrogen sources can be provided in a range of states of reduction (e.g. ammonium, nitrate, urea), where the resulting nitrogen metabolism can have dramatic effects on the media pH through secretion or uptake of protons [15]. Small-scale culturing typically obscures the buffering role of CO2 because the amount of gas supplied is in great excess of stoichiometric requirements, such that the CO2 approaches an equilibrium condition that is amenable for algae growth [14]. In outdoor operation, pH control is often implemented by sparging CO2 on demand (CoD) when the pH goes above a specified upper bound, and the resulting carbonate equilibrium acidifies the medium [1]. CO2(g) → CO2(aq) + H20 ↔ H+ + HCO− 3

J. Wang et al. / Algal Research 18 (2016) 288–295

Because CO2 suffers from poor solubility of gasses in warm water, the majority of the CoD may not transfer into the culture for consumption, thereby representing a significant process cost as well as greenhouse gas emissions. Attempting to retain gas while allowing light penetration rapidly results in rapid heating of the culture and/or significant evaporative losses. Rather than utilizing CO2 buffering, acids or bases can be added via feedback control based on pH sensor measurements. This approach is hindered, however, by the undesirable accumulation of inhibitory counter-ions and costs of acids—which are nearly as great as the nitrogen sources [6]. The conceptually simpler and more economical approach of formulating balanced growth-media by combining ammonium and nitrate nitrogen sources is limited due to the significant and preferential assimilation of ammonium before nitrate by algae [12], which lowers pH to the point of cell death. In a circumstance where the algae cannot regulate its own nitrogen utilization to maintain pH, the alternative is to supply different nitrogen sources based on the metabolism that will result in the desired pH control. Metabolic modeling-based pH control faces many challenges that are even greater than control issues in a typical bioreactor due to the dynamic condition of outdoor culture; one must consider both predictable and unpredictable components of light, temperature, rainfall, and evaporation rates—all operating on different time scales. While incorporating sophisticated feed-forward modeling is a long-term goal that combines both the biology, physics, and environment, this initial manual implementation was focused on translating laboratory experience to outdoor conditions—including brutal temperatures exceeding 40 °C to gain further insights into the overall challenge. The DOE's ATP3 (Algae Testbed Public-Private Partnership) facility at Arizona State University's (ASU) Mesa campus' AzCATI (Arizona Center for Algae Technology and Innovation) provided an opportunity for access to a large number of analytics and PBRs designed specifically for established algal cultivation methods. The transition from indoor experimental studies [14] to outdoor operation in this work represents a useful characterization of what challenges to anticipate in the next phase of moving the proposed metabolic pH control strategy from pilot to large scale systems. In this work, a simple stoichiometric-based metabolic model for open-loop pH control was implemented based on daily predicted nitrogen demand (PND) and ammonium addition into nitrate-based algae medium in an outdoor pilot-scale, flat-panel PBR at AzCATI. The existing method of CO2 on demand (CoD) was operated as the control. The observations and impact of this approach on operation and performance of these two different pH control operational strategies are discussed.

289

prepared by adding stock solutions to tap water inside the bubble-columns and PBRs. 2.1.3. Inoculation and cultivation scale-up Microalgae were pre-cultured indoors and scaled up sequentially for inoculation of 60 L flat-panel PBRs. Microalgae were initially cultured in bubble-columns first at 0.1 L and then 0.8 L working volume. They were then inoculated at a minimum OD750 of 0.3 and cultured in logarithmic phase at 10–15 L working volume in 15 L flat-panel PBR with dimensions of 4ʹ × 4ʹ × 0.05′. All indoor cultures were cultivated under continuous light at 200–300 μE m−2 s− 1 light intensity supplied by polychromatic fluorescent light, while receiving 2% (v/v) CO2 gas. Outdoor PBRs (Fig. 1) were then inoculated at a minimum OD750 of 1.0 (to avoid photo-bleaching) at 50 L working volume in flat panel PBRs with dimensions of 48ʺ × 72ʺ × 4ʺ supplied by gas bubbling from the bottom via 1/16″-diameter holes drilled on the side of a PVC pipe. Prior to inoculation, the PBRs were thoroughly cleaned with a water jet, then filled with a 0.03% NaClO bleach solutions for at least 30 min contact time, and rinsed 3 times with water to remove residual chlorine levels as confirmed using test strips Pooltime™ 6-way test strips. The temperature range for indoor bubble columns and PBRs was 22– 28 °C based on laboratory room thermostat. For outdoor PBRs, culture temperature ranges of 24–29 °C were maintained by cycling 25 °C chilled water from a nearby Glacier® pool evaporative cooler model #GPC-250 through 1 cm ID stainless steel cooling coil with a 6 m of contact length per PBR. Temperature and pH of outdoor PBR cultures were measured using a handheld pH 100 ExStick® meter and probe. The pH probe was calibrated with temperature compensation using standard pH 4, 7, and 10 buffer solutions and was recalibrated every night to adjust for signal drift during cultivation. Data was acquired by YSI 5200 A multiparameter monitor, wirelessly transmitted, and recorded to a PC

2. Materials and methods 2.1. Algae cultivation methods 2.1.1. Microalga strain The marine microalgae Nannochloropsis oceanica sp. 0209 was obtained from the National Center for Marine Algae and Microbiota, Bigalow, Maine (formally known as the National Center for Culture of Marine Phytoplankton (CCMP). This alga was chosen as the model marine, lipid accumulating eukaryotic microalgae species for the comprehensive multi-site evaluation of algae performance for the U.S. Department of Energy ATP3 program (www.atp3.org). 2.1.2. Growth media Cultures were maintained in f/2 salt-water medium. Modified f/2 medium consisted of (per 1 L): 0.754 g NaNO3, 0.067 g NaH2PO4, 0.0063 g FeCl3·6H2O, 0.002 g Na2EDTA·2 H2O, 1.8 × 10− 4 g MnCl2·4H2O), 6.3 × 10−6 g Na2MoO4·2H2O, 9.8 × 10−6 g CuSO4·5H2O), 1.0 × 10−5 g CoCl2·6H2O), 2.2 × 10−5 g ZnSO4·7H2O), and 35 g Oceanic ™ Sea Salt Mix (Detroit, MI, etc.). A stock solution of 500 g/L NH4Cl was prepared for pH adjustment for outdoor PBRs. Culture media were

Fig. 1. Flat panel photobioreactor used for outdoor experimentation at the Arizona Center for Algae Technology and Innovation (AzCATI).

290

J. Wang et al. / Algal Research 18 (2016) 288–295

computer running AquaManager software version 6.0 Build 5001 housed in a nearby control room. A web browser-based virtual control panel is also available through the on site wireless network using a smart phone. 2.2. Photobioreactor sampling and analysis Approximately 1 L of culture was sampled daily at 12:00 noon from the outdoor PBRs and analyzed for (a) dry weight (DW), (b) ash-free dry weight (AFDW), (c) fatty acid methyl ester (FAME) content; (d) substrate concentration of ammonium and nitrate in supernatant (e) optical density at wavelength 750 nm (OD750) and 680 nm (OD680). Also, approximately 20 mL of culture was sampled from the outdoor PBRs immediately before, 10 min after, and 2–5 h after dosing NH4Cl and analyzed for (a) OD750 and OD680 and (b) substrate concentration for ammonium and nitrate. 2.2.1. Optical density, dry weight and ash free dry weight OD750 and OD680 were measured using the Hach® DR 2800 v.1.11 bench-top spectrophotometer with non-linear calibration and 1 mm quartz/plastic cuvette. Algal culture samples were blanked with f/2 medium, diluted to an OD of 0.2–0.8, and read in triplicate at both OD750 and OD680, which further served to quickly indicate culture stress levels. Total dry algae biomass and dry ash were determined in triplicate using a gravimetric filtration and oven drying. A measured volume of sample visually determined to contain at least 5 mg of dry biomass was added to pre-ashed and pre-weighed filters in manifolds while applying a vacuum pressure. The filter with algae cake was then rinsed with 1 M ammonium formate solution to remove salt dried in an oven for 12 h at 105 °C, cooled in a desiccator for 1 h, and then re-weighed to determine combined total dry algae biomass and ash. The filter was then ashed in a muffle furnace at 500 °C, cooled in a desiccator for 3 h, and re-weighed to determine total ash. 2.2.2. Light availability as assessed by integration of photosynthetically active radiation (PAR) Due to outdoor growth, light levels vary throughout the day. While a rigorous assessment of light flux to the reactor requires a projection of the solar beam onto the photobioreactor geometry, we present a simpler approach to reasonably assessing light availability by simply integrating the light normal to the earth as measured by a field-mounted LICOR Biosciences LI-250 A light meter with quantum Q41000 PAR sensor (μE/m2/s).

process Z time

Cumulative PAR ¼

PARðt Þ  dt inoculation

In this manner, the differences in light of a given day can be captured while still accommodating for lack of light at night and effectively replacing time as the dependent variable with a quantitative index of light availability. This approach to presenting the data also provides for comparisons of growth in artificial constant light environments or with defined variable photoperiods. In these outdoor studies, the vertical line of midnight encompasses the entire dark culture period. 2.2.3. Substrate concentration of nitrate and ammonium Substrate concentration of nitrate, ammonium, and orthophosphate in culture samples were determined using QuickChem® methods with references for nitrate + nitrite [16] and for ammonium [5]. Samples are flow-injected in triplicate into a Lachat Instruments QuikChem® 8500 Automated Ion Analyzer equipped with Omnion 3.0 software for analysis.

2.2.4. Elemental analysis of biomass Elemental composition for carbon and nitrogen in algal biomass was measured with a CE Instruments Elemental Analyzer EA 1110 (Thermo Electron Corp). For this, about 2–5 mg of biomass was lyophilized, homogenized, and instantly combusted into CO2 and NOx, gaseous products which were then collected and analyzed to estimate mass composition of carbon and nitrogen. 2.2.5. Alkalinity Total alkalinity is the acid equivalents (equ) required to neutralize dissolved media species to a final pH of 3.5. 50 mL of algae culture supernatant was sampled and stored at 4 °C prior to measurements. The sample was incrementally titrated with 0.02 N H2SO4 solution while monitoring the pH from the initial condition to a final pH of 3.5 (corresponding to a complete removal of media bicarbonate). 2.2.6. FAME lipid analysis Analysis of Fatty Acid Methyl Ester (FAME) content is performed with Agilent Technologies GC analytical instrumentation following standard basic transesterification procedure developed in NREL for algae biomass [8]. 2.3. Photobioreactor pH control strategies 2.3.1. Feedback pH Control using CO2 sparging CoD based pH control utilized automatic CO2 sparging to reduce the pH. The upper pH setpoint was programmed as 7.9 (lower bound = 7.8, upper bound = 8.0), where a normally-closed solenoid would open to deliver CO2 at a flow rate of 12.5 L min− 1 (LPM) (or 24.6 g min−1) until pH reached the lower bound, though the preset sample interval of 15 min often resulted in the pH well below the lower bound. 2.3.2. Feed-forward pH control using metabolic assimilation of nitrogen PND based pH control utilized a sparge of CO2 just sufficient for inorganic carbon demand such that pH could be controlled by the proton imbalance associated with nitrogen consumption. The constant supply flow rate of 0.25 CO2 LPM was empirically determined to provide growth rate-dependent proton imbalance, determined as a moderately fluctuating pH as described in more detail below. Growth stoichiometry provides a well-defined quantity of protons according to: αCO2 þ ψi Ni þ δH2 0→CHx Ny Oz þ λO2 þ Φi Hþ

ð1Þ

where (ψ/ϕ)NH4 ≈ 0.92 and (ψ/ϕ)NO3 ≈ −1.12. Therefore, there is a direct correlation between the amount of growth and the anticipated secretion or uptake of protons as developed previously [15]. ΔHþ metabolism;i ¼ ðψ=ΦÞi •ξλ ðΔODλ Þ

ð2Þ

where ξλ is the wavelength (λ)-dependent conversion factor between optical density (OD) and cell concentration (X) (e.g. OD750 = ξ750•X )[10]. A nitrate-based basal media was used for cultivation, and NH4Cl was dosed at appropriate concentrations and time intervals as described below. The constant supply flow rate of 0.25 CO2 LPM was iteratively determined to be the CO2 supply rate needed to sustain growth at a relatively constant rate while minimizing pH fluctuation (±0.01 pH change per 15 min). A feed-forward manual dosing for NH4Cl was calculated based on the anticipated growth for the next culture period utilizing the stoichiometry represented in Eq. 2. Predicting the amount of NH4Cl to add for the next day was straight-forward due to linear growth, a result of light limitation and minimal light disruption in the summer Arizona climate. Note that although any growth kinetic model could be utilized to calculate the predicted increase in optical density (ΔOD750,predicted), we utilized an experimentally determined average dry weight conversion factor (ξ750 = 0.3) under the initial

J. Wang et al. / Algal Research 18 (2016) 288–295

assumption of a typical algae biomass composition of 10% N dry weight. ΔðNH4 ClÞpredicted ¼ ðψÞNH4Cl •ξ750 ΔOD750;predicted

ð3Þ

Note that these initial approximations for conversion factors are readily updated and refined based on experimental yield and experimentally determined Nannochloropsis biomass composition. An important consideration for experimental execution is how large a daily NH4Cl addition will be, and into how many increments the feed must be divided to provide for a reasonably small change in pH as a result of proton secretion during metabolism to amino acids. Based on preliminary experiments, the anticipated growth capability in the PBR corresponds to an ΔOD750 ≤ 2 with a corresponding biomass accumulation of 0.5 gDW/L. Based on 10% nitrogen content by weight, this corresponds to a maximum nitrogen assimilation of 0.05 g N/L (3.6 mM NH4-N). Near neutral bioreactor pH, this level of secretion of protons would correspond to a capacity for unbuffered pH change of 4.6 pH units (ΔpHunbuffered = − 4.6), therefore, to maintain a pH change of a pH unit due to secreted protons would require 4–5 incremental media additions per day. It is important to note, however, that in the case of a salt-water based media, the pH change due to alkalinity is significantly buffered near neutrality and is on the order of 0.5 ΔpH units/mN-equivalents H+ (as will be confirmed from an alkalinity titration curve, Fig. 5). A secretion of 3.6 mM H+ would therefore correspond to a pH change of 1.8 pH units (ΔpHsea-water = − 1.5), suggesting that two media additions should be sufficient to maintain the ammonia-metabolism based pH change within a given day to less than a single pH unit. In the actual experiment, based on the end of day sample from previous day, the growth predictions (OD750) are: ΔOD750,day3 = 2.06, ΔOD750,day4 = 1.61; ΔOD750,day5 = 1.38, and calculated NH4Cl dosages (mmol) are 4.40, 3.45, and 2.95, respectively. The preceding calculations provide a basis for experimental design as well as insight into the fundamental principles that govern pH control (i.e. considerable dependence upon the dynamics of nitrogen utilization, CO2 supply). Due to inherent constraints of ATP3 outdoor configurations, CO2 sparge rate could not be readily varied to minimize use and match stoichiometric carbon demand. In this study, the PBR was therefore initially operated at constant CO2 supply rate without NH4Cl dosing for the first two days to generate a baseline CO2 supply and associated pH fluctuation as well as provide culture growth rates from which to predict the subsequent NH4Cl supplements. The daily dosage calculated by Equation 3 was split evenly into two doses administered at 9:00 (a time in the day when culture was expected to have already photo-adapted) and 15:00 (a time in the day to allow for the assimilation of the majority of the first dose and provide ammonium for the second half of the day during maximum PAR).

291

accuracy. An ANOVA confirmed the obvious linearity of the dry weight and cumulative PAR time courses, and a Student t-test compared the linear slopes for bioreactor correlations between dry weight, ash-free dry weight and cumulative PAR and found no statistical difference at a 99.9% confidence level. The standard deviations of the slopes were calculated from matrix operations in Excel. Although a piecewise linear model could be used for ANOVA comparison of provided and summed analytical nitrogen measurements, the large differences obviate the utility of a statistical assessment. 3. Results and discussion 3.1. Growth for alternative pH control strategies Both flat-panel photobioreactors operated using the different pH control strategies, respectively PND (metabolic pH control using predicted nitrogen demand) and CoD (CO2 on-demand pH control), displayed similar growth (Fig. 2). PND and CoD both displayed an OD750 which increased from 2.5 to 9 with only minimal biomass loss during the unlighted night-time hours. The appearance of the PND suggested greener culture which is reflected in the optical density (OD) ratio of OD680/OD750 which correlates to the chlorophyll-a content to dry weight. This ratio has also been suggested as an indicator of the level of stress within a culture [2]; in this case however, since the metabolic pH control is mediated by increased levels of nitrogen addition in the form of ammonium, this likely reflects the impact of that additional nitrogen on chlorophyll content [10] and not a specific physiological stress. Additional indicators of comparable growth are shown as more direct measurements of dry weight and associated ash content. In Fig. 3, the daily measurements of dry weight, %-ash and associated ash-free dry weight (AFDW) are shown with respect to the integrated photosynthetically active radiation (PAR) as described in the methods. In this particular outdoor study, since the solar insolation is quite constant

2.4. Replication and statistical analysis Biomass composition CHN analysis was measured in triplicate for all samples with a consistent standard deviation of 0.06%N and 0.09%N respectively for CoD and PND photobioreactors. Media nitrate and media ammonium ion concentrations were performed in triplicate for all samples with an average analytical standard deviation of 0.06 ppm for nitrate and 0.08 ppm for ammonium (corresponding to error bars smaller than figure datapoints). The absolute variation duplicate total FAME measurements was on average 0.37% FAME for CoD photobioreactor, and 0.25% FAME for PND photobioreactor (smaller than figure datapoints); the relative reproducibility for overall FAME measurements is a standard deviation of 2.6% of the measured values. Optical density was taken as single values due to insignificant differences for replicate measurements. Dry weight and ash free dry weight for each sample was taken as single measurements to provide sufficient material for accurate gravimetric measurement. The accuracy of continuous monitored data such as pH and PAR relied on daily calibration for

Fig. 2. Growth of flat panel PBRs grown with alternative pH control strategies of CO2 addition on demand (CoD) and fixed minimal CO2 addition with pH control based on predicted nitrogen metabolic demand (PND). A) growth as indicated by OD750 (light scattering regime) and B) an indicator of chlorophyll pigmentation from the ratio of absorbance max for chlorophyll-a (680 nm) to the biomass level (750 nm). Timepoints indicated correspond to midnight.

292

J. Wang et al. / Algal Research 18 (2016) 288–295

Fig. 3. Biomass dry weight and ash content as a function of cumulative photosynthetically active radiation (PAR) for Nannochloropsis grown in flat panel PBRs with alternative pH control strategies of CO2 addition on demand (CoD) and predicted metabolic nitrogen demand (PND). A) Dry weight and ash free dry weight (AFDW) and B) %-ash content over a 5-day outdoor growth at AzCATI site.

(within 5% for the 5 day period, data not shown), the relatively consistent sampling time and light conditions for these biomass measurements results in very little need for correction to PAR relative to process time. We feel that it is important, however, to present data in the context of solar insolation as it provides a more accurate basis for comparing data not only across different growth periods, climates, and sites, but also to indoor growth conditions, which are often operated with continuous light. The observation of a nearly perfectly linear increase in biomass over these five days strongly suggests that the growth under both PBR conditions is light-limited and the slightly increased chlorophyll content was not significant in terms of improving biomass accumulation. These correlations were found to be linear by ANOVA (p b 0.006) and the slopes of the two bioreactors were not statistically different (α b 0.001) for a ash free dry weight yield of 7.0 ± 0.2 mg/L per PAR (mole photons per aerial square meter). There is a small but consistent variation in %-ash that could reflect small changes during the increase in biomass loading, but is more likely a systematic error in the measurement of these quantities. 3.2. Culture pH and alkalinity The two alternative pH control methodologies resulted in very different pH profiles with the CoD fluctuating rapidly between a pH of 8 and 6.5 and the PND displaying relatively smooth variations from 7.5 to as high as 9 (Fig. 4A,B). Comparable growth despite these large pH changes corroborates the general observation that many algae have a rather broad pH growth range [7] which provides flexibility in implementing pH control while trying to minimize nutrient resources. The outdoor pilot scale setting presented a considerable challenge to maintaining pH control for both methods as neither approached the optimal pH control achievable in small indoor systems. An ammonium addition twice each day provides minimal control relative to prior work with multiple small additions under constant indoor conditions [12]. Similarly, the 0.25 vvm sparge rate of the CO2 pH control line was observed to be excessive by the formation of large bubbles, achieving poor CO2 transfer. In combination

Fig. 4. Time courses of pH, temperature and solar insolation (PAR) for the alternative pH control strategies. A) pH fluctuations during metabolic demand for nitrogen assimilation (PND). B) pH fluctuations during CO2 sparging on demand (CoD) C) Temperature in both reactors, and D) solar insulation measured as photosynthetically active respiration (PAR).

with a slow sampling rate (i.e. 15 min. to accommodate highly multiplexed logging), this resulted in comparatively large pH fluctuations relative to what can be achieved with fine bubble dispersion, rapid sampling, and more precise gas flow control. Ambient daytime temperatures exceeding 40 °C presented a physical challenge to personnel, and resulted in bioreactor temperatures that varied from 20 to 30 °C throughout the day and night despite the internal PBR heat exchangers. These fluctuations substantially alter gas solubility as well as pH probe response despite daily efforts to maintain calibration with off-line measurements and temperature compensation. The comparable rates of growth observed for these highly variable conditions are a testament to the resilient nature of algae, which is further reflected by its growth kinetics in the light-limited growth condition, where the effects of other factors are attenuated by comparison. The fluctuations in temperature result from the daily dynamics of light flux (Fig. 4D), where it should be noted that at ‘solar noon’, a vertical flat panel will have a period of reduced light due to the sun's rays predominantly impinging on the narrow top, rather than the sides of the photobioreactor. This results in the ‘bimodal’ response of pH and corroborates the choice for sufficient but not excessive gas flows. During the reduced photon flux of ‘solar noon’, the reduced CO2 consumption rate is reflected in a temporary drop in pH, wherein CO2 instead favors dissolution into and subsequent acidification of the media, until the light starts to enter the other side of the flat panel reactor where higher CO2 consumption rates resume and increase the pH. The baseline flow rates of 0.005 vvm CO2 utilized during the PBR operation resulted in an estimate of 25.8% efficiency for carbon utilization based on the experimental average 45% carbon dry weight basis. Although this reflects a relatively low efficiency of gas-liquid mass transfer, it is an order of magnitude higher than what is typically observed in heterotrophic bioreactors for oxygen. A poor efficiency of gas use is typical for bioreactors where for example O2 use in heterotrophic microbial culture also utilizes b5% of the supplied gas. By reducing gas flow and not relying on CO2 for pH control the supply can be much closer to the demand. Alkalinity encompasses a measure of the media capacity to neutralize acid, which was measured on samples at noon the day after a

J. Wang et al. / Algal Research 18 (2016) 288–295

previous afternoon inoculation where the media alkalinity of both flat panel PBRs were comparable (CoD, 6.51 mEq/L; PND, 5.79 mEq/L) with a slight elevation in alkalinity for photobioreactor that was more heavily sparged with CO2. The alkalinity for metabolic control (PND) was found to drop slightly by the end of culture to 5.73 mEq/L in contrast to the CO2 sparge approach which increased substantially to 8.49 mEq/L over the 5 day culture period (Fig. 5A). The alkalinity titration curves measured at the end of culture are provided in Fig. 5. As expected, the use of CO2 sparging as the basis of suppressing pH results in an increase in alkalinity relative to metabolic pH control with reduced CO2 buffering. Although increased alkalinity enhances pH buffering capacity, which functions as passive control, it makes active model predictive control (MPC) of pH response to metabolic proton flux more difficult. To quantify this behavior, the derivative slope of the alkalinity curves can be compared at pH = 6. The more highly buffered CoD bioreactor displays a derivative of −2.0 mM H+/ ΔpH as compared to −4.44 mM H+/ΔpH, which are both dramatically smaller than an unbuffered water system (− 2.33 × 106 mM H+/ ΔpH). Implementation of model-predictive pH control would require

Fig. 5. Alkalinity ranges during growth of Nannochloropsis in flat panel photobioreactors with alternative pH control strategies of CO2 addition on demand (CoD) and predicted nitrogen metabolic demand (PND). The lower panel presents the alkalinity titration curve which provides a direct illustration of the effects of alkalinity on process control because the slope of the alkalinity is a measure of buffering capacity and attenuates the change in pH for a give secretion or uptake of protons due to nitrogen metabolism.

293

this information either as empirical experimental input or a more complex prediction of the buffering based on media species. It is noteworthy that the metabolic control (PND) treatment did not only accumulate acid-neutralizing species, but also demonstrated minimal reduction in the alkalinity, suggesting that any contribution of ammonium metabolism to the neutralization of media, was minor relative to buffering provided by the saltwater and the baseline CO2 sparge rate. As noted in the background, the ability to achieve metabolic pH control is dependent upon the growth stoichiometry providing sufficient protons from ammonium assimilation. The total ammonium-N fed to the PND was 130 mg N/L (9.3 mM NH4-N/L), which corresponds roughly to the same acid equivalents based on growth stoichiometry. Combined with the observed slope of the titration curves in the relevant operational range of pH ~6.0 of 0.5 ΔpH-units/H+ mEq the capacity for pH change in this experimental implementation was 5 pH units. Since the ammonium nitrogen is comparable to the initial nitrate ammonium (8.9 mM NO3-N/L) there is comparable potential for pH control based on secreted and assimilated protons. Noting that the total growth of the two PBRs was comparable, it is unclear how much of the nitrogen was assimilated which emphasizes the critical nature of verifying the overall nitrogen mass balance. In this particular implementation, the overall biomass yield on nitrogen for the PND was reduced due to starting with the same media nitrate content in both photobioreactors. An improved strategy for the nitrogen assimilation pH control method would start with much lower nitrate levels with fed-batch addition to minimize total nitrogen fed and allowing for controlled nitrogen limitation. Since the methodologies of nitrogen measurement have undergone considerable standardization at the AzCATI site, it is informative to examine the nitrogen mass balance by adding up the total of the various measurements including nitrogen supplied, inside the cells, and within the media. As shown in Fig. 6, the media nitrate levels (measured at noon) had declined to essentially zero for both bioreactor configurations. The ammonium levels were functionally zero throughout the entire run, even in the case of PND where NH4Cl was added roughly three hours before the mid-day sample point. The biomass nitrogen composition measured between 5 and 7.5% by weight, declined after nitrate depletion as expected for the CoD bioreactor and remained between 6 and 7% for the CoD case that included ammonium feed (Fig. 6A). The lower panel (Fig. 6D) represents a predicted total nitrogen mass balance as compared to the sum of the various analytically measured components multiplied by the magnitude of that nitrogen pool (total cells and total media). The sum of the analytical methods is considerably lower than predicted by mass balance, suggesting there is either an unaccounted medium nitrogen-containing species or a systematic error associated with the measurements and/or the associated calculations. The daily ammonia volatilization rate was measured to be around 3 mg/L/day (~0.5% per day of f/2 media N) from a separate experiment utilizing an un-inoculated PBR, suggesting that this is not a significant contributor to nitrogen loss. While secretion of amino acids or proteins remains a possibility, the levels of unaccounted nitrogen are quite large and seems unlikely. Noting that the inorganic nitrogen media components contribute a declining and eventually negligible amount to the overall mass balance, the associated errors must come from the elemental analysis and dry weights. The batch nitrogen addition of the CoD is more easily interpreted, since the total nitrogen (corrected for media and cell sampling) should remain constant. The constantly declining summed total nitrogen indicates that the biomass composition analysis or its associated dry weight determination is likely systematically underestimating the actual nitrogen levels. This is consistent with a cumulative loss of calculated nitrogen as well, since successive samplings would be removing more nitrogen than calculated in a systematic cumulative error in the overall mass balance. The deviation of the PND which includes nitrogen addition twice a day does display the trend of a calculated increase in total system nitrogen, but that amount falls considerably short of what is predicted by mass balance. There are several

294

J. Wang et al. / Algal Research 18 (2016) 288–295

Fig. 7. Fatty Acid Methyl-Esters (FAME) lipid profiles for the two outdoor flat panel PBRs (m = metabolic controlled PND, c = feedback CO2 controlled, CoD) give the fatty acid compositions throughout the 5-day growth period for Nannochloropsis.

4. Conclusions

Fig. 6. Assessments of nitrogen mass balance during operation of the flat plate photobioreactors using ‘standard’ methods for media and cellular nitrogen content. A) Biomass nitrogen content (% of dry weight), B) Media nitrate levels, C) Media ammonium levels (near zero), and D) a comparison of predicted and summed total media an cell content of nitrogen reflecting a ‘loss’ of considerable nitrogen due to sump of the individual analytics.

important implications to this observation. First, the tendency to utilize data as absolute quantities and even conversion factors can result in considerable error. Second, the goal of model predictive control is undermined by systematic inaccuracies in the associated nitrogen mass balance. Successful implementation of the proposed strategy requires a greater focus on validating the overall mass balance of nutrients and avoiding the typical focus on reproducibility that can miss large systematic errors.

Utilizing CO2 to both provide carbon and media buffering results in poor biomass yield on CO2 and a large loss of the provided CO2 into the environment. Choosing a low baseline flow of CO2 that allows pH fluctuations combined with ammonium additions to introduce proton secretion provides a means to decrease CO2 usage and still maintain pH in a range for growth. The alkalinity presented by sea-water and increased by CO2 represents a significant challenge to implementing nitrogen-metabolism based pH control due to the reduction in pH response caused by alkalinity. Nonetheless, the slope of titration curves in the range of desired operating conditions combined with the stoichiometrically released protons due to growth gives a pH response that was sufficient to implement control. Achieving such a control will require models of growth that can more precisely predict the nitrogen mass balance as well as growth in response to the dynamic conditions of outdoor growth. The largely manual implementation of sampling, nitrogen demand prediction and feedings is not sufficient to achieve the desired control, and will require more sophisticated instrumentation and control algorithms. Most importantly, the nature of high density algae growth which is inherently light-limited, constrains the response of the system to be surprisingly predictable despite the large fluctuations in light and pH observed in these outdoor studies. This insensitivity of growth response should be an asset towards implementing a pH control that is more precise, and includes manipulation of CO2 sparge rates for example to further reduce the overall CO2 use during algae photobioreactor operation. This type of bioreactor control work requires conducting such studies in the challenging environment of outdoor algae culture systems. Contributions

3.3. Lipid productivity Significant differences in fatty acid profiles were observed for the alternative pH control strategies (Fig. 7). The CoD culture had final lipid content that was 8% higher lipid than that of PND. This was most likely caused by the depletion of nitrogen in CoD which did not occur in PND because it was periodically dosed with an ammonium that effectively doubled the total N content of the bioreactor. The increased lipid content was comprised primarily of C16:0 (palmitic acid), C16:1n7 (palmitoleic acid), and C18:1n9 (oleic acid), the major fatty-acid constituents of Nannochloropsis oceanica lipid fractions [3,4]. The content of high-value ω-3 fatty acid C20:5 (eicosapentaenoic acid [EPA]), however, remained constant, as previously observed in literature [11]. Therefore, in this specific case, nitrogen starving might not have been a determinant for EPA productivity.

All authors approved the final version of the submitted manuscript. JM provided funding/facilities for this work. JW designed the study, acquired and analyzed data, and drafted the article. TR and PW provided technical support and acquisition of data. WRC conceptualized the study and provided critical analysis/interpretation of the data in manuscript revisions critical for important intellectual content. Acknowledgements This work was funded in part through the Department of Energy's Bioenergy Technologies Office (BETO), Award Number: DEEE0005996, the Algae Testbed Public-Private Partnership (ATP3) - a RAFT Partnership. The experiment and data analysis also received significant assistance from analytical and production specialists of the

J. Wang et al. / Algal Research 18 (2016) 288–295

Arizona Center for Algae Technology and Innovation at Arizona State University, managed by Thomas Dempster with particular note for biochemical analysis by Maria Bautista and Sarah Kempkes, and the cultivation set-up and operations by Mark Segel, Nicholas Csakan and David Cardello of the laboratory of Dr. Peter Lammers. We thank Erica Lennox at the Curtis Laboratory for editing assistance and critical review of the manuscript. References [1] M. Berenguel, et al., Model predictive control of pH in tubular photobioreactors, J. Process Control 14 (4) (2004) 377–387 Available at: http://www.sciencedirect.com/science/article/pii/S0959152403000738 [Accessed April 4, 2012]. [2] K.A. Chekanov, A.E. Solovchenko, Possibilities and limitations of non-destructive monitoring of the unicellular green microalgae (Chlorophyta) in the course of balanced growth, Russ. J. Plant Physiol. 62 (2) (2015) 270–278 (Available at:) http:// link.springer.com/10.1134/S1021443715010033. [3] G. Chini Zittelli, et al., Production of eicosapentaenoic acid by Nannochloropsis sp. cultures in outdoor tubular photobioreactors, Prog. Ind. Microbiol. 35 (C) (1999) 299–312. [4] B. Crowe, et al., A comparison of nannochloropsis salina growth performance in two outdoor pond designs: Conventional raceways versus the arid pond with superior temperature management, Int. J. Chem. Eng. (2012). [5] D. Diamond, Determination of ammonia (Phenolate) by flow injection analysis, In QuikChem®Method 10–107–06-1-I, Loveland, CO., Lachat Instruments, 1995. [6] N.E. Khan, et al., A process economic assessment of hydrocarbon biofuels production using chemoautotrophic organisms, Bioresour. Technol. 172 (2014) 201–211 (Available at:) http://www.sciencedirect.com/science/article/pii/S0960852414012346 (Accessed September 8, 2014).

295

[7] H. Khatoon, et al., Effects of different salinities and pH on the growth and proximate composition of Nannochloropsis sp. and Tetraselmis sp. isolated from South China Sea cultured under control and natural condition, Int. Biodeterior. Biodegradation 95 (2014) 11–18 (Available at:) http://linkinghub.elsevier.com/retrieve/pii/ S0964830514002042. [8] L.M.L. Laurens, et al., Accurate and reliable quantification of total microalgal fuel potential as fatty acid methyl esters by in situ transesterification, Anal. Bioanal. Chem. 403 (1) (2012) 167–178. [9] J. Liu, et al., Freshwater microalgae harvested via flocculation induced by pH decrease, Biotechnol. Biofuels 6 (1) (2013) 98 (Available at:) http://www. biotechnologyforbiofuels.com/content/6/1/98. [10] J.A. Myers, B.S. Curtis, W.R. Curtis, Improving accuracy of cell and chromophore concentration measurements using optical density, BMC Biophys. 6 (1) (2013) 4. [11] D. Pal, et al., The effect of light, salinity, and nitrogen availability on lipid production by Nannochloropsis sp, Appl. Microbiol. Biotechnol. 90 (4) (2011) 1429–1441. [12] M.L. Scherholz, W.R. Curtis, Achieving pH control in microalgal cultures through fedbatch addition of stoichiometrically-balanced growth media, BMC Biotechnol. 13 (2013) 39. [13] R. Slade, A. Bauen, Micro-algae cultivation for biofuels: cost, energy balance, environmental impacts and future prospects, Biomass Bioenergy 53 (0) (2013) 29–38 Available at: http://dx.doi.org/10.1016/j.biombioe.2012.12.019. [14] J. Wang, Improving Process Efficiency of Algae-Based Biofuel and Biproduct Production Using Metabolism-Based pH Control. The Pennslyvania State University, pH.D. THesis, Dept of Chemical Engineering, 2015 (127 pp). [15] J. Wang, W.R. Curtis, Proton stoichiometric imbalance during algae photosynthetic growth on various nitrogen sources: toward metabolic pH control, J. Appl. Phycol. 28 (1) (2015) 43–52 (Available at:) http://link.springer.com/10.1007/s10811-0150551-3. [16] K. Wendt, Determination of nitrate/nitrite in surface and wastewaters by flow injection analysis, In QuikChem®Method 10–107–04-1-A, Loveland, CO., Lachat Instruments, 1995.